Try it out and see how it does for you. If you're willing to share its report on power, I'll give you free credits.
Interesting read. I genuinely worry all the time about Canada falling behind with AI. It seems so hard to turn around and the US absolutely has its own incentives.
Very happy to be presenting on forecasting research results, based on the Social Science Prediction Platform data, at the IDB's EconNet seminar series.
There's a public Zoom which you can find by googling, in case it's of interest (trying to avoid bots).... It starts at 12 ET!
Just a note that with all this "LLMs accurately identifying users from text", some free text responses in allegedly de-identified survey data could in principle identify research respondents, if not today then soon.
If you use any of my work in your teaching, please let me know so I can cite it in my tenure package! Thanks!
(We have a pretty late deadline here.)
Pretty cool project!
Did you ever wish you could get help with your power calculations? Now you can!
earlyreview.ai will give you estimates of treatment effects based on your early project documents (e.g., pre-analysis plans, registered reports, grant proposals, etc.).
^ The above table is based on work with the Social Science Prediction Platform. How human forecasts can also improve power calculations is in the paper: www.nber.org/papers/w34493
I recently presented some results that showed that using LLM forecasts in experimental design could dramatically improve power.
Studies in our data start severely under-powered (~0.4). Using LLM forecasts would bring it up to ~0.7! And further improvements coming.
This is from the co-founder of Anthropic, interesting that he refers to public sources when he is also obviously privy to lots of internal sources that he cannot discuss. I assume he sees the same thing at Anthropic.
importai.substack.com/p/import-ai-...
